[Intel-gfx] [PATCH v5 01/13] drm/i915/icl: Configure lane sequencing of combo phy transmitter

[Lisovskiy, Stanislav]

On Mon, 2018-09-10 at 20:57 +0530, Madhav Chauhan wrote:
> On 9/10/2018 5:50 PM, Lisovskiy, Stanislav wrote:
> > On Tue, 2018-07-10 at 15:10 +0530, Madhav Chauhan wrote:
> > > This patch set the loadgen select and latency optimization for
> > > aux and transmit lanes of combo phy transmitters. It will be
> > > used for MIPI DSI HS operations.
> > > 
> > > v2: Rebase
> > > 
> > > Signed-off-by: Madhav Chauhan <madhav.chauhan at intel.com>
> > > ---
> > >   drivers/gpu/drm/i915/icl_dsi.c | 38
> > > ++++++++++++++++++++++++++++++++++++++
> > >   1 file changed, 38 insertions(+)
> > > 
> > > diff --git a/drivers/gpu/drm/i915/icl_dsi.c
> > > b/drivers/gpu/drm/i915/icl_dsi.c
> > > index 13830e4..a571339 100644
> > > --- a/drivers/gpu/drm/i915/icl_dsi.c
> > > +++ b/drivers/gpu/drm/i915/icl_dsi.c
> > > @@ -105,10 +105,48 @@ static void gen11_dsi_power_up_lanes(struct
> > > intel_encoder *encoder)
> > >   	}
> > >   }
> > >   
> > > +static void gen11_dsi_config_phy_lanes_sequence(struct
> > > intel_encoder
> > > *encoder)
> > > +{
> > > +	struct drm_i915_private *dev_priv = to_i915(encoder-
> > > > base.dev);
> > > 
> > > +	struct intel_dsi *intel_dsi = enc_to_intel_dsi(&encoder-
> > > > base);
> > > 
> > > +	enum port port;
> > > +	u32 tmp;
> > > +	int lane;
> > > +
> > > +	/* Step 4b(i) set loadgen select for transmit and aux
> > > lanes
> > > */
> > > +	for_each_dsi_port(port, intel_dsi->ports) {
> > > +		tmp = I915_READ(ICL_PORT_TX_DW4_AUX(port));
> > > +		tmp &= ~LOADGEN_SELECT;
> > > +		I915_WRITE(ICL_PORT_TX_DW4_AUX(port), tmp);
> > > +		for (lane = 0; lane <= 3; lane++) {
> > > +			tmp = I915_READ(ICL_PORT_TX_DW4_LN(port,
> > > lane));
> > > +			tmp &= ~LOADGEN_SELECT;
> > > +			if (lane != 2)
> > > +				tmp |= LOADGEN_SELECT;
> > > +			I915_WRITE(ICL_PORT_TX_DW4_LN(port,
> > > lane),
> > > tmp);
> > > +		}
> > > +	}
> > > +
> > > +	/* Step 4b(ii) set latency optimization for transmit and
> > > aux
> > > lanes */
> > > +	for_each_dsi_port(port, intel_dsi->ports) {
> > > +		tmp = I915_READ(ICL_PORT_TX_DW2_AUX(port));
> > > +		tmp &= ~FRC_LATENCY_OPTIM_MASK;
> > > +		tmp |= FRC_LATENCY_OPTIM_VAL(0x5);
> > > +		I915_WRITE(ICL_PORT_TX_DW2_AUX(port), tmp);
> > > +		tmp = I915_READ(ICL_PORT_TX_DW2_LN0(port));
> > > +		tmp &= ~FRC_LATENCY_OPTIM_MASK;
> > > +		tmp |= FRC_LATENCY_OPTIM_VAL(0x5);
> > > +		I915_WRITE(ICL_PORT_TX_DW2_GRP(port), tmp);
> > > +	}
> > > +}
> > > 
> > 
> > I think bspec states that latency optimization should be set only
> > for
> > Transmit lanes 0, 1, 3. Is it fine to use a group access(i.e
> > ICL_PORT_TX_DW2_GRP) here? I think it states also that no latency
> > optimization is needed for the clock lane.
> 
> There is a separate comment added in BSPEC :
> "The Latency Optimization of the Clock Lane can be either left at
> it's 
> default value ('h0)
> or programmed to the same value as the other lanes. If programmed
> with 
> the same
> value as the other lanes,  then the Group access can be used for 
> PORT_TX_DW2 programming"

Yep, I saw that, however just wasn't sure did they mean we actually
should leave it untouched. Is there actually any difference, if it is
programmed or not? If the value is discarded anyway, I guess it should
have been stated explicitly. Nevermind then.

> 
> Regards,
> Madhav
> 
> > 
> 
> 
-- 
Best Regards,

Lisovskiy Stanislav